Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Thesis submitted to the faculty of the Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of
Master of Science
in
Mining and Minerals Engineering
ABSTRACT
The potential for improved selectivity has made column flotation cells a popular choice
for upgrading fine coal. Unfortunately, recent production data from full-scale column
plants indicate that many industrial installations have failed to meet original expectations
in terms of clean coal recovery. Theoretical studies performed using a simple dispersion
data showed that this low-cost modification increased flotation recovery as predicted by
the dispersion model. This study presents the key findings obtained from the field
investigation and provides generic guidelines for designing multi-stage column circuits.
TABLE OF CONTENTS
iii
2.1.2 Commercial Installations .............................................................................. 15
iv
LIST OF FIGURES
Figure 2. Conventional (left) and column (right) flotation cell installations. ..................4
Figure 4. Entrainment of fine mineral matter with the water reporting to the
Figure 5. Comparison of how feed water reports to the froth in mechanical and
Figure 7. Effect of residence time on recovery for different feed solids contents
Figure 10. Effect of Sb on flotation recovery for two different air sparging
v
Figure 11. Column with deaeration tank used to control froth handling problems. ........30
Figure 13. Recovery as a function of rate (k), residence time (τ) and Peclet
Figure 14. Graphical solution used to find the Peclet Number from RTD data. .............36
Figure 15. Visualization of column mixing using a red dye and timed-sequence
Figure 16. Visualization of column mixing using a red dye and timed-sequence
Figure 17. Comparison of RTD data for single- and two-stage column circuits. ............41
Figure 18. Recovery curve for projecting two- and three-stage column circuit
performance... .................................................................................................42
Figure 19. Column circuit layout (a) before and (b) after cell-to-cell
reconfiguration. ..............................................................................................44
Figure 20. Production statistics before and after the cell-to-cell column
reconfiguration. ..............................................................................................45
Note: Unless otherwise noted, all images are the property of the author.
vi
LIST OF TABLES
Table 1. Example of kinetic parameters for a typical coal flotation system. .................11
Table 3. Example of RTD data for a four-cell bank of conventional machines. ...........35
Table 4. Example of RTD data for single- and two-stage column circuits. ...................43
Table 5. Production statistics before and after the full cell-to-cell column
reconfiguration. ................................................................................................46
vii
ACKNOWLEDGMENTS
The author would like to acknowledge several people who have helped make this
thesis possible. First and foremost, thanks and appreciation is given to Dr. Gerald H.
Luttrell for his continued guidance, assistance, and friendship. The author would also
like to thank Dr. Greg T. Adel and Dr. Roe-Hoan Yoon for their advice and wisdom
The author would like to thank the many individuals at Alpha Natural Resources,
such as Fred Stanley, Frank Addison, and Paul King, for their help and support
A heartfelt thank you goes out to the author’s best friend and spouse, Jane
Kennedy, for her love, companionship, and support. The author would also like to convey
his deepest gratitude and appreciation to the rest of his family for their unconditional love
and encouragement.
viii
1.0 INTRODUCTION
1.1 Background
1.1.1 Preamble
concentrating finely ground ores. The basis of this process is the utilization of the
differences in the natural or created hydrophobicity of the two minerals that are to be
separated. The separation involves chemical treatment of an ore pulp to create conditions
favorable for the attachment of selected mineral particles to air bubble. The combined
air/mineral particles have specific gravities less than the pulp and will float to the surface,
forming a mineral laden froth that can be skimmed off from the flotation unit while the
the most efficient, most complicated, most sensitive, most challenging and least
understood mineral processing operation today. The art of flotation has been correctly
chemistry. Only in the last three decades has much progress been made in understanding
its mechanisms.
Over 95% of all base metals are concentrated by flotation. In coal preparation
plants using flotation, approximately 5-15% of the raw coal feed is processed through the
flotation unit for recovery of fine coal particles. Virtually the entire free-world supply of
copper, lead, zinc, and silver is recovered in the froth of flotation machines. As the grade
(the assay of valuable constituent) of mineral deposits decreases and coal and metal
1
prices rise, the importance of flotation increases. In turn, the necessity of improving
separation of different particles can be achieved when air is injected into the slurry. The
basic difference between ore and coal flotation is that for ores the entire tonnage is
ground to flotation size, whereas for coal only the fine fraction (about 60 mesh x 0) not
In coal flotation, the differences in the wettability of coal (hydrophobic) and rock
(hydrophilic) are exploited. Hydrophobic particles strike bubbles and attach (see Figure
1), while hydrophilic particles do not attach. This attachment occurs when air is injected
into the slurry containing coal and rock. The coal particles attach to the air bubbles and
Froth
Feed
Tails
2
are buoyed to the pulp surface. The rock particles remain in the slurry and are rejected as
tailings. The pulp phase provides an opportunity for particles to selectively attach to the
air bubbles. The froth phase is used to separate bubble-particle aggregates from the pulp.
The last objective is to form at the surface of the cell a semi-stable mineral-laden
froth that is fragile enough to allow entrapped rock materials to fall and stable enough to
retain the mineral particles selectively. The froth must also have sufficient strength to
permit mechanical or natural removal from the machine. Once the froth is out of the
flotation cell, it should break down rapidly to allow pumping of the concentrate. The
product”.
There are two basic types of flotation machines and adaptations of these machines.
Conventional (mechanical) and column flotation cells are the two types of flotation
flotation machines. Conventional cells are open top vessels connected in rows (banks)
through which the pulp flows continuously from the head (feed) end to the tail end. These
machines are typically arranged in banks of 4-6 cells. The cells are individually agitated
using an impeller. The impeller provides the energy needed to keep particles in
suspension, suck in air (a blower may also be used) and to disperse air into small air
bubbles.
3
Figure 2. Conventional (left) and column (right) flotation cell installations.
A column cell consists of a vertical column into which air is added at the bottom
using spargers (no impeller). Columns usually use wash water to prevent fine clays from
being entrained into the clean coal froth. Gas compressors are usually required to provide
the air flow and energy for air dispersion. A column cell is usually a stand-alone machine
reagents (modifiers). Coal flotation is primarily concerned with collectors and frothers.
Collectors are the heart of the flotation process since they are the reagents that produce
the hydrophobic film on the mineral particle. Coal, although hydrophobic, is enhanced by
the collector increasing the rate at which it reports to the froth phase and helps the coal
4
attach to the air bubble well. The collector should be applied as far back in the flotation
circuit as possible because conditioning time with the flotation feed is very important.
The conditioning time becomes even more important with harder to float coal such as low
rank bituminous or oxidized coal. Conditioning time in a flotation circuit may range from
seconds to minutes.
The most common types of collectors for coal flotation are diesel fuel, kerosene,
and fuel oil. There are also environmentally friendly collectors available such as bio-
diesel and tree products. The collector dosages range from 0 to 5 lb/ton. Higher rank
bituminous coal may require no collector, low rank bituminous may require as much as 3
lb/ton of coal and weathered (oxidized) coal may require 5 lb/ton of coal.
The main purpose of the frothing agent is the creation of a froth capable of
carrying the mineral-laden bubbles until they can be removed from the flotation machine.
the bubble. The life of the individual bubble is thus prolonged until it can be further
stabilized by adherence of mineral particles and joined with other bubbles at the pulp
surface to form a froth. Once it is withdrawn from the flotation machine, however, the
froth should break down rapidly, to prevent interference with subsequent processing
operations. The frother decreases the size of the bubble in the pulp, thus increasing the
surface area of the air introduced into the flotation machine. This enlarges the number of
collisions of the mineral particles with the bubble and chance of the mineral to be floated
The ability to lower the surface tension of water is a characteristic of all frothers.
Since most organic compounds can do so to at least a limited degree, the number of
5
frothers commercially available might be expected to be quite large. However, the
following requirements on the frothing reagent have restricted the frothers to only a few
• be low in cost,
• be readily available,
• be effective in low concentrations,
• have no collecting properties or have collecting affinity to the valuable mineral,
• be non-toxic,
• not have a repulsive odor,
• produce a persistent froth that will last the length of the flotation bank, but that
will collapse when skimmed into the launder,
• be stable during storage even at adverse atmospheric conditions (low and high
temperatures),
• have strength enough to withstand the turbulence in the flotation cells and to carry
the valuable mineral, and
• be brittle enough to allow the rejection of unwanted gangue material.
Frothers may consist of short-chain alcohols, derivatives, glycols, and glycol mixtures.
The dosages of frother for conventional cells should typically be in the range of 7 to 15
ppm. The dosages for a column cell, which are usually higher, should typically be from
10 to 25 ppm.
Flotation performance is dependent on rate constant (k), retention time (τ), and
cell mixing (Pe). The interrelationship between these variables is described in greater
detail later in this document. Retention time is the amount of time the coal particle has to
stay in the pulp. In a coal flotation circuit, 3½ to 4 minutes of residence time is needed
for good recoveries. Mean retention time can be determined by using retention time
in which V is the active volume of the cell and Q is the flow rate to the cell. The number
and size of the flotation cells should be adjusted to have the proper amount of time for
flotation. RTD testing involves using a tracer solution, such as salt, applied to the feed
end of a flotation cell and taking timed samples vs. concentration of each sample at the
Cell mixing is the last but not least of the areas affecting flotation. The best
flotation recoveries would come from a plug flow system rather than a well mixed
system. Conventional cells are set up to try and obtain plug flow using cell to cell.
The rate constant (k) indicates how fast coal floats and is dependent on coal
type/size/solids content/feed rate/froth depth, reagent type/dosage, gas flow rate and
sparging/agitating system. Some of the factors that can impact flotation kinetics are
described below.
Floatability is very dependent on the rank of coal. Low volatile, mid volatile, high
volatile, anthracite, or oxidized (weathered) coals all float differently. Low volatile coal
is the easiest to float and oxidized coal is the hardest to float. Oxidized coal causes
problems with poor recoveries even when blended with higher rank coals.
consistency within the 60 mesh x 0 is very important to coal flotation. The courser the
7
coal the harder it is to keep attached to the air bubble. The finer the coal is will lead to
The solids content of a flotation circuit should be around 3-7% depending on the
classification system. At higher the solids contents, a trade-off between residence time
Froth depth changes how wet or dry the froth will be for a given reagent dosage.
A deep dry froth reduces entrainment and clean coal ash. However, a very deep froth may
become unstable and will lower recovery. The type of frother should be tested to
determine the best product for each flotation circuit. Dosages of frother should be
determined to optimize the flotation circuit. The types and dosages of frother can increase
The gas flow rate should be sufficient to provide the bubble surface area needed
to carry out the particles. Conventional cells should use about 3.5 cfm/ft2 and columns
about 5.0 cfm/ft2. The rate the coal floats decreases with low gas flow rates.
moving for attachment to the air bubble. The energy supplied to the flotation unit should
8
1.3 Flotation Modeling
Despite being a complex process, flotation can often be modeled using simple
kinetic expressions. For a batch flotation cell, recovery can be predicted based on:
R = 1-exp(-kτ) [2]
in which R is the factional recovery, k is the rate constant (1/min) and τ is the mean
retention time (min). Flotation feeds are typically assumed to contain three components:
Therefore, mathematically, total recovery (Rtot) can be calculated from the factional
recoveries of fast (Rf), slow (Rs) and non-floating (Rn) components using:
Rtot = Rf + Rs + Rn [3]
Rf = φf (1-exp{-kfτ}) [4]
Rs = φs (1-exp{ksτ}) [5]
Rn = φn (1-exp{-knτ}) [6]
in which φf, φs and φn, are the respective mass fractions of fast, slow and non-floating
solids and kf, ks and kn are the corresponding rate constants of fast, slow and non-floating
function of time (see Figure 3(a)). The froth products collected from this procedure
include both floatable solids as well as entrained solids. Release analysis, on the other
9
Feed
Products Collected
Clean #1 (1/2 min) As a Function of
Cumulative Time
Clean #2 (1 min)
Clean #3 (2 min)
Clean #4 (4 min)
Step I Step II
Feed
Separation of Separation of
Nonfloatables Clean #1 Floatables
Clean #2
Clean #3
Clean #4
Tail #1
b) Release Test
Tail #2
hand, is often used to establish the best possible separation results for flotation. It is used
for plant design and for assessing flotation efficiency. As shown in Figure 3(b), the
process of running a release analysis consists of two steps. The first step is to separate the
floating from the non-floating coal. Step two the fast floating coal to slow floating coals
are separated into grades and results are calculated to give grade recovery data and plots
associated with yield and recovery. The final froth products collected from this procedure
include only floatable solids since entrained solids have been removed upstream. Table 1
shows a representative set of kinetics data for a typical coal flotation system.
10
Table 1. Example of kinetic parameters for a typical coal flotation system.
method for upgrading coal fines that are too small to be treated using density-based
separators. Most of the installations in the coal industry use mechanical (conventional)
allows some ultrafine impurities (e.g., clay) to be hydraulically carried into the clean coal
product in the process water that reports to the froth. This contamination substantially
reduces the overall quality of the clean coal froth product. The entrainment problem has
forced many coal operations to install column flotation cells in place of conventional
flotation machines (Davis et al., 1995). Column cells are able to reduce entrainment by
the addition of a countercurrent flow of wash water to the top of the froth. The water
rinses ultrafine impurities such as clay from the froth back into the flotation pulp. Studies
indicate that less than 1% of the feed slurry reports to the froth in a properly operated
column. However, a recent field survey showed that while column technology can indeed
improve clean coal quality, many industrial installations are unable to provide good coal
11
Since recent production data from full-scale column plants indicate that many
industrial installations have failed to meet original expectations in terms of clean coal
recovery, a detailed field study was undertaken at an industrial site (i) to identify the root
cause of the problem responsible for the poor coal recovery in column cells and (ii) to
develop low-cost methods by which this problem may be minimized. The study included
mixing using a simple dispersion model. This document presents the key findings
obtained from the field study and provides generic guidelines for designing multi-stage
column circuits. The resultant data is used to demonstrate the large financial impacts of
12
2.0 LITERATURE REVIEW
2.1.1 Overview
The major problem with conventional flotation is that it allows a fine mineral
(clay) slimes to be recovered with the water reporting to froth (see Figure 4). Column
cells can overcome this shortcoming by washing the clay slimes from the froth product
using wash water (see Figure 5). Sufficient wash water must be added to ensure that all of
the feed water that normally reports to the froth product has been replaced with fresh or
clarified water. The froth depth must also be much greater than that used in conventional
flotation to obtain good distribution of the wash water and to achieve the desired froth
AIR
Attachment
Entrainment
AIR
Coal AIR
Middling
Mineral
Figure 4. Entrainment of fine mineral matter with the water reporting to the froth
product.
13
Froth
Mechanical Column Water
Flotation Froth Flotation
Water
Wash
Water
Feed
Feed Water
Water
Tailing Tailings
Water Water
Figure 5. Comparison of how feed water reports to the froth in mechanical and
column flotation cells.
washing action. This unique feature makes it possible for column cells to achieve
superior ash removal when compared to conventional flotation cells. In most cases, the
separation performance obtained using column cells approaches very closely to the
Although column cells offer superior performance in terms of ash removal, many
operators in the coal industry have been reluctant to adopt the column technology. The
comparatively low market value of fine coal often makes it difficult for mine
management to justify the higher capital and operating costs for columns, particularly if
the expenditure is for the replacement of existing conventional cells. Many operators also
have the perception that columns are more difficult to operate, entail greater amounts of
maintenance, and require complicated ancillary systems for compressed air and wash
14
water. Nevertheless, several operators have identified suitable niches where the
are currently in commercial service in the U.S. coal industry. As shown, the most popular
brands of columns include Microcel, Eriez/CPT (CoalPro) and Jameson. Although the
Jameson cell is not really a “column”, it is included in this list since it typically uses wash
water to improve ash rejection. Details related to the specific design features of the
15
various column technologies are available in the literature (McKay et al., 1988; Yoon et
al., 1992; Davis, 1995; Manlapig et al., 1993; Finch and Dobby, 1990; Finch, 1995;
Rubinstein, 1995; Wyslouvil, 1997). Due to economy of scale, recent trends in the coal
industry have shifted toward the installation of smaller numbers of very large (>14 ft)
diameter units. Although most of the installations involve the treatment of minus 100
mesh feeds, some operations have been able to justify the use of columns to treat
deslimed 100x325 mesh streams. In such cases, the columns have been used to further
The primary advantage of column cells is their ability to achieve a higher level of
and column flotation cells at an industrial plant site. The column data were obtained
using both a laboratory (2 inch diameter) test column and a full-scale (10 ft diameter)
production column. For comparison, test data were also obtained from a single bank of
five full-scale Denver cells (300 ft3 each) already in operation at the preparation plant. A
release analysis test was also conducted on the flotation feed to provide a baseline for
comparative purposes (Dell et al., 1972). Release analysis is often described as the
Several important conclusions may be drawn from the results given in Figure 6.
First, the separation data obtained using the column cells fall along nearly identical
16
100
Release
Analysis
60
40
Full-Scale Conventional
20 Laboratory Column
Full-Scale Column
0
0 5 10 15 20
recovery-grade curves regardless of column size. This finding demonstrates that the data
obtained from the smaller test columns can be used to accurately predict the performance
of full-scale columns.
The scale-up of separation data obtained using conventional cells is generally less
reliable due to unpredictable variations in froth product quality. These variations are
drainage behavior for conventional cells of different sizes (Lynch et al., 1981). Second,
the performance data for the column cells tend to fall just below the separation curve
predicted by release analysis. This finding implies that the column cells provide a level of
17
performance that would be difficult to achieve even after multiple stages of cleaning by
conventional machines. For this particular example, the column cells reduced the
concentrate ash by nearly 8 percentage points compared to the existing plant rougher
All flotation cells must have adequate slurry residence time in order to achieve an
acceptable recovery of clean coal. The mean residence time can be calculated by dividing
the active volume of the flotation pulp by the volumetric flow rate of refuse. For a typical
100 mesh x 0 feed, a residence time of 3.5-4.0 minutes is often adequate for 3-5 cell bank
machines minimizes the bypass of feed slurry to tailings. In contrast, column cells are
usually arranged in parallel, which can allow some floatable coal to bypass to tailings. As
a result, columns typically require significantly more residence time to achieve the same
level of coal recovery (Dobby and Finch, 1985; Mankosa et al., 1992). This situation is
particularly serious for the relatively low profile columns that are typically employed in
the coal industry This problem can lead to lower recoveries of coal if not accounted for
The detailed procedure for estimating the column residence time required to
achieve a given recovery has been presented in the literature (Dobby and Finch, 1986a).
This procedure is complicated and well beyond the scope of this article. However, as a
rough rule-of-thumb, a column cell will typically require approximately twice the
residence time of a 4-cell bank of conventional cells and three times the residence time of
a batch laboratory flotation cell. Therefore, it would not be unusual for a column cell to
18
require 7-8 minutes of residence time. Some industrial columns have been designed to
operate with residence times of 12-15 minutes or more. It is also important to note that
the active volume for columns is defined as the volume of pulp contained between the
bottom of the froth and the top of the air spargers. The calculated volume is also typically
reduced by 10-20% to account for the volume occupied by air bubbles. This volume is
substantially less than the total volume of the column tank structure.
time for a coal flotation column. The volumetric feed flow rate to the column was
steadily increased in this series of tests in order to reduce the mean residence time. As
shown, the test data obtained at 2% solids correlated well with the theoretical
performance curve predicted using Levenspiel’s (1972) equation. However, as the solids
content of the feed stream was increased to 5% solids, the recovery dropped sharply and
deviated substantially from the theoretical curve. This deviation was more pronounced
when the solids content was increased to 10% solids. The drop in recovery with
capacity. When this occurs, there is insufficient bubble surface area to carry all of the
floatable particles through the froth. While the carrying capacity restriction can often be
column cells. This is largely due to the fact that the specific surface area of the cell (ratio
of the cross-sectional area to the volume) is much higher for conventional cells.
19
100
Theoretical
2% Solids
80 5% Solids
10% Solids
Recovery (%)
60
40
20
Figure 7. Effect of residence time on recovery for different feed solids contents
(Luttrell et al., Proceedings of the 21st Annual International Coal,
Aggregrate and Mineal Processing Exhibition and Conference, 2004).
Theoretical studies indicate that carrying capacity, which is normally reported in terms of
the rate of concentrate floated per unit cross-sectional area, is linearly related to the size
and density of particles in the froth (Espinosa-Gomez et al., 1988; Sastri, 1996). The froth
carrying capacity may be estimated from laboratory and pilot-scale flotation tests by
conducting experiments as a function of feed solids content (Finch and Dobby, 1990).
Most of the full-scale columns in the coal industry operate at carrying capacities in the
20
range of 0.06 to 0.24 tph/ft2, with an average of about 0.12 tph/ft2 for 100 mesh x 0 feeds.
The values at the lower end of this range typically correspond to coal feeds of much finer
particle size (325 mesh x 0). Once this value is known, the required column cross-
sectional area can be determined by dividing the expected clean coal tonnage (tph) by the
convenient to calculate the clean coal tonnage for a full-scale column from a smaller test
column using:
2
L arg e Column tph ⎛ L arg e Diameter ⎞
=⎜ ⎟ [7]
Small Column tph ⎜⎝ Small Diameter ⎟⎠
According to this relationship, a 14 ft diameter column should have a clean coal capacity
that is nearly twice that of a 10 ft diameter column [i.e., (14/10)2 = 1.96]. Field data
suggest that this relationship holds valid for coal columns as large as 15 ft diameter.
stating that froth mobility problems make overflow lip length (and not cross-sectional)
This expression may be more appropriate for columns that operate with unstable froths
feeds.
The most important feature of a column is the wash water system. A froth depth
of about 2-3 ft is typically required to ensure good distribution of the wash water and to
21
prevent short-circuiting. In addition, the flow of wash water must exceed the volumetric
flow of water reporting to the clean coal product to prevent entrainment of the high-ash
slimes. In most cases, less than about 1% of the feed water will report to the froth product
if the wash water is properly controlled. The amount of water carried by the froth can be
calculated from:
⎛ 100 ⎞
Q = 4 C⎜ − 1⎟ [9]
⎝ P ⎠
Where Q is the water demand (gpm/ft2), C the froth carrying capacity (tph/ft2) and P is
the froth percent solids. For example, a column cell producing 0.12 tph/ft2 of dry clean
coal at 18% solids will carry about 2.2 gpm/ft2 of water from the pulp into the froth [ i.e.,
washes (defined as the froth water demand divided by the wash water addition rate)
reaches a value of one. However, as shown in Figure 8, froth mixing usually requires that
1.5 dilution washes be used to fully suppress hydraulic entrainment. This constraint
dictates that a wash water flow rate of at least 3.3 gpm/ft2 [i.e., 1.5 x 2.2 = 3.3] be used in
the current example to prevent the entrainment of high-ash slimes. Field data collected
obtained from columns operating in the coal industry suggest that a wash water flow rate
higher gas and frother addition rates will typically increase the froth water demand and,
as a result, the amount of wash water required. Excessive wash water flows should be
avoided since the extra wash water passing downward through a column will create an
undesirable reduction in the slurry retention time and, hence, a potential reduction in
22
Figure 8. Effect of number of dilution washes on froth quality (Luttrell et al.,
Proceedings of the 21st Annual International Coal, Aggregrate and
Mineal Processing Exhibition and Conference, 2004).
recovery. Very high water additions may also destabilize the froth by stripping surfactant
(frother) from the bubble surfaces. High water rates may also have a detrimental impact
on product grade by increasing axial froth mixing and reducing the wash water
The design of the wash water distributor can also significantly affect column
performance. In some cases, the distribution piping is intentionally submerged below the
cell dip so that a drained froth can form above the distributor. This arrangement allows
23
the depth of the drained froth and the extent of froth drainage to be varied by raising or
lowering the distributor. Changes to the vertical position of the water distributor can be
used to somewhat control the split of water between the clean coal and refuse streams. In
some cases, multi-level concentric distribution rings may also be used to overcome
problems associated with poor froth mobility. The inner rings are typically located above
the outer rings to reduce drainage and improve the fluidity of the froth in the center of the
column. In other cases, the water distributor may be located just above the top of the
froth. This arrangement does not allow the froth mobility to be controlled by adjusting
the distributor location, but it does make if easier to identify and correct plugging
problems that may severely impact the performance of the distribution network.
The air sparging system is the most important, and perhaps most controversial,
component in the design of a coal flotation column. Ideally, the spargers should produce
small, uniformly sized bubbles at a desired aeration rate. The spargers should also be
non-plugging, wear-resistant and allow for easy, on-line servicing. Three primary types
• porous bubblers fabricated from filter cloth, punctured rubber tubes or sintered
columns), and
24
Details related to these systems have been discussed extensively in the technical literature
(Dobby and Finch, 1986b; Xu and Finch, 1989; Huls et al., 1991; Yoon et al., 1992;
popular column spargers. In general, the data for all three systems tended to fall along
nearly identical separation curves. This suggests that the selectivity of the separation is
independent of the method used to produce the air bubbles. Similar data have been
100
Combustible Recovery (%)
80
60
40
Porous Bubbler
20 Air Injector
Static Mixer
0
0 20 40 60 80 100
selection is unimportant. The test data indicate that the porous bubbler and static mixer
generally produced higher overall recoveries than the high-velocity air injector. In
addition, the static mixer was generally less sensitive to changes in operating conditions
and tended to produce data points higher on the separation curve. On the other hand,
systems such as the porous bubbler have been shown to plug easily and require frequent
maintenance in industrial operations. Capital costs for the static mixer system are
generally higher due to the purchase of a slurry pump and wear can become an issue
when very coarse feeds (>100 mesh) are treated. All these factors need to be carefully
bubble surface area generated by each device. This capability is commonly reported in
terms of the superficial bubble surface area rate (Sb), which is defined as the total bubble
surface area per unit of time passing through a given column cross-sectional area. This
Sb = 30.5 V g/ Db [4]
in which Vg is the superficial gas rate (cfm/ft2) and Db is the bubble diameter (mm). The
superficial air rate may be calculated by dividing the volumetric gas flow rate (cfm) by
the column cross sectional area (ft2). The impact of Sb on flotation recovery is illustrated
by the test results given in Figure 10. These data show that recovery increases sharply as
Sb increases above 50 sec-1 and eventually reaches a plateau at 100-150 sec-1. These
26
100
90
70
60
50
40
30
20 Porous Bubbler
10 Static Mixer
0
0 50 100 150 200
Superficial Bubble Area Rate (1/sec)
Figure 10. Effect of Sb on flotation recovery for two different air sparging systems
(Kohmuench, Personal Communication, 2000).
values indicate that nearly 1 million square feet bubble surface area passes through a 14 ft
diameter column every minute! In addition, the data indicate that the relationship
between recovery and Sb is generally independent of sparger type. This finding implies
that most commercial sparging systems can achieve a target recovery provided that a
It should also be noted that operation at Sb values above 150 sec-1 often produces
poor separations because of “runaway” froths that cannot be effectively washed. In most
industrial cases, a superficial gas rate in the range of 3.5-5.0 cfm/ft2 would be suitable for
27
coal applications. These values correspond to total aeration rates of 540-770 scfm of air
for a full-scale 14 ft diameter column. The gas rates at the lower end of the range would
generally be used for spargers that generate smaller bubbles, while the higher gas rates
are typically needed for less efficient spargers. A proper combination of gas rate and
bubble size will generally provide a gas holdup in the flotation pulp in the range of 15-
18%. The holdup can be determined by installing pressure transducers at two different
levels along the height of the flotation column. A low air holdup indicates that the
Caution should also be used during the metering of gas flow rates. A properly
designed system should be equipped with a flow meter that is calibrated to read correctly
placing a pressure regulator ahead of the flow meter. By placing the air flow control
valve after the flow meter, the flow meter will always operate at its design pressure. If the
flow meter is placed after the control valve, then the operating pressure and true gas flow
rate are both unknown. Improper metering of the gas flow rate can be a particularly
serious problem when laboratory and pilot-scale tests are conducted for the purpose of
for column applications. In fact, several installations of columns in the coal industry have
required the purchase of additional compressors to reach the original design flow rate.
Much of this confusion is related to improper use of gas flow terminology (Sullair
Bulletin, 1992). For example, column manufacturers normally report gas flow rates as a
“standard” volumetric flow per time. This value is only valid at 1 atm (14.7 psia) of
28
pressure and 20oC (68oF) of dry air. The “actual” flow rate specified by compressor
this amount of air enters the compressor, it is not necessarily the amount of air delivered
to the column due to compressor seal leakage. As a result, the “actual” flow may be only
95% of the “inlet” flow. Furthermore, corrections to the gas flow rate must be made to
account for differences in elevation (atmospheric pressure) and humidity. Air temperature
generally has little impact on the capacity of an oil-flooded screw compressor, but may
require that professionals be consulted to ensure that the compressor is properly sized for
the specified air requirements. This item is not an issue with the Jameson cell that
Concentrates containing large amounts of ultrafine (325 mesh x 0) coal generally become
reducing frother dosage have not been successful and generally lead to lower column
recoveries. Therefore, several circuit modifications have been adopted by the coal
industry to deal with the froth stability problem. For example, column launders need to be
considerably oversized with steep slopes to reduce backup. Horizontal froth travel
distances must be kept as short as possible, while adequate vertical head must be
provided between downstream operations and the column launder. Most of the newer
column installations include a deaeration tank to permit time for the froth to collapse (see
29
Figure 11). Special provisions are also normally required to ensure that downstream
dewatering units can accept the large froth volumes. Standard screen-bowl centrifuges
equipped with 4 inch inlets may need to be retrofitted with 6 inch or larger inlets to
minimize flow restrictions. Some installations may require defoaming agents to deal with
the froth stability problem. These reagents can represent a large operating cost and must
Deaeration
Tank
Column
Cell
Spargers
Figure 11. Column with deaeration tank used to control froth handling problems.
30
2.3 Residence Time Modeling
ensure that all coal particles have the opportunity to collide and adhere to air bubbles.
This contacting time is commonly referred to as the flotation time or residence time. As a
coals. The required residence time may be even longer for difficult-to-float coals that are
very fine or superficially oxidized. The mean residence time (τ) can be estimated by
dividing the active cell volume (V) by the volumetric flow rate (Q) of slurry passing
τ =V /Q [10]
Note that the active volume of the cell discounts the volume of the impeller mechanism
and any hold-up of gas due to rising air bubbles. For example, a bank of four 14.2 m3
machines (85% active volume) fed 817.6 m3/hr of slurry would have τ≈3.5 minutes (4
cells x 14.2 m3/cell x 0.85 / 817.6 m3/hr x 1 hr/60 min = 3.5 minutes). While this rule-of-
thumb is useful for a first-pass evaluation, the specification of a minimum residence time
The mean residence time represents the average length of time that it takes an
element of fluid to travel through the flotation cell. In a plug-flow or batch system, all
fluid elements are exposed to the same residence time (Figure 12). Likewise, in a
perfectly-mixed system, some elements begin to exit immediately after being introduced
31
due to the perfect dispersion. However, in a continuous bank containing partially mixed
or more than one agitated tank, a wide distribution of residence times may exist for each
element of fluid entering the system. The differences in this dispersed-flow system are
allow some of the flotation pulp to exit the machine at a time different from the mean
residence time.
may be obtained for the same mean residence time (Finch and Dobby, 1990). This
by Wehner and Wilhelm (1956), which states that the total recovery (R) for any first-
4 A exp{Pe / 2}
R = 1− [11]
(1 + A) exp{( A / 2) Pe} − (1 − A) 2 exp{(− A / 2) Pe}
2
Concentration
Concentration
Concentration
where τ is the mean residence time, k is the flotation rate constant, and Pe is the Peclet
number. For flotation systems, the terms τ and k are usually expressed in units of minutes
and 1/minutes, respectively. As a result, Equation [11] can be plotted as a function of two
The product kτ represents the degree of completion of the kinetic process, while
parameter Pe represents the intensity of dispersive mixing. The operating band shown in
Figure 13 is bounded by the upper plug-flow (Pe=∞) curve and the lower perfectly-mixed
(Pe=0) curve. While a plug-flow system is obviously more desirable since it provides a
much higher recovery than a perfectly-mixed system at the same kτ, plug-flow conditions
100
90
Plug-Flow
Recovery (%)
80
4 8
70 2
1
0.5
60
Perfectly
Pe
50 Mixed
40
0 1 2 3 4 5 6
Figure 13. Recovery as a function of rate (k), residence time (τ) and Peclet number
(Pe).
33
are difficult to duplicate in industrial practice. As a result, most commercial flotation
machines are designed with multiple agitated tanks arranged in a cell-to-cell bank in an
attempt to approach plug-flow behavior. A simple calculation shows that the total
RN = Ri + Ri (1 − Ri ) + Ri (1 − Ri ) 2 + Ri (1 − Ri ) 3 + ... + Ri (1 − Ri ) N = 1 − (1 − Ri ) N [6]
where Ri is the fraction recovery defined at τi=τ/N. For a bank of perfectly-mixed cells,
N
⎛ N ⎞
RN = 1 − ⎜ ⎟ [13]
⎝ N + kτ ⎠
As expected, the performance curve provided by this equation approaches that of a plug-
typically limited to four or five tanks in series since additional cells provide little
Equation [11] is often inadequate to describe the kinetic response of individual flotation
cells. Instead, most industrial machines operate with 0<Pe<∞. Therefore, Pe must be
determined to establish the appropriate recovery versus kτ curve for industrial flotation
34
(i) adding a tracer substance to tag an incoming element of fluid to a flotation
machine and
time.
An example of RTD data for a four cell bank of conventional cells is shown in Table 2.
The mean residence time (τ) can be calculated from these data using (Levenspiel, 1972):
τ=
∑ tCΔt [14]
∑ CΔt
where C is the tracer concentration at any time t and Δt is the time increment between
samples [Δt= (ti+1+ti-1)/2]. This procedure gives τ=3.47 minutes for the data provided in
Table 3. Pe is related to the spread of the RTD, which can be quantified by the
35
dimensionless variance (σθ2) given by (Levenspiel, 1972):
σθ 2
=
∑ t CΔt − 1 = 2⎛⎜ 1 − 1 − e
2 − Pe
⎞
⎟⎟ [15]
τ ∑ CΔt
2 ⎜ Pe 2
⎝ Pe ⎠
numerical iteration. For convenience, the graphical solution is plotted in Figure 14 for the
determined from the RTD data in Table 3. Since Pe is known, Equation [11] can now be
Figure 14. Graphical solution used to find the Peclet Number from RTD data.
36
3.0 EXPERIMENTAL RESULTS AND DISCUSSION
Conventional flotation machines are typically arranged in series with the tailings
from each cell providing feed to the next. This multi-stage arrangement makes it possible
for conventional machines to maintain good coal recoveries even at relatively short
residence times. Column cells, on the other hand, are usually fed in parallel and operate
independently from each other. This layout dictates substantially more residence time be
provided for columns to maintain good recoveries of floatable particles (Nicol and
Bensley, 1988). However, even at long residence times, the lack of downstream
scavenging creates a potential for feed to bypass to tailings in a relatively short time.
column flotation cell, a series of tests were conducted using colored dye (dark red) to
visually observe the mixing behavior of fluid passing through a clear Plexiglas flotation
cell. The 30-inch diameter cell was filled with fluid to a height of 60 inches to provide a
The flow rate of gas to the cell, which was equipped with a MicrocelTM sparging, was
held constant at 2.0 cm/sec for all tests. A feed flow rate equivalent to a superficial
velocity of 0.25 cm/sec was maintained by injecting fresh water through a 4-way feed
distributor placed near the top of the liquid level. This feed flow rate provided a total
volumetric residence time of approximately 10 minutes. However, when corrected for air
hold-up, the mean residence was probably in the range of 7-8 minutes. The transport of
37
the dye through the cell was observed at one-second intervals using a digital color camera
Figure 15 shows the results obtained from the first test conducted using the tracer
dye. The first image (at 1 sec) is located in the top-right corner of the composite
photograph, while the last image (at 16 secs) is located at the bottom-left corner. The
13 9 5 1
14 10 6 2
15 11 7 3
16 12 8 4
Figure 15. Visualization of column mixing using a red dye and timed-sequence
images at one-second intervals.
38
timed sequence of images shows that the dye is first visible through the column wall at
about 4 seconds. Surprisingly, the dye appeared to be thoroughly mixed throughout the
entire cell volume after only about 13-14 seconds. This finding suggests that column cells
Contributors to the high degree of internal mixing are likely to include the small length-
to-diameter ratio and the agitation and circulation provided by the sparger pump.
Another series of tests were conducted using a column cell equipped with baffles
designed to minimize mixing. The baffles consisted of four horizontal partitions that split
the cross-sectional of the column into four “pie” slices of equal area. The baffles passed
from just below the top of the liquid level, down pass the the sparger inlets, and
terminated just above the central tailings discharge. Unfortunately, the test data showed
that the dye appeared to be mixed throughout the cell volume after only about 15-16
seconds (Figure 16). In addition, the data showed that the partitions tended to provide
unequal distribution of the feed in each quadrant, which would be an undesirable result
for industrial operations. Therefore, the qualitative data obtained in this study suggests
that this type of baffling cannot effectively minimize large internal mixing.
RTD measurement was conducted at an industrial column flotation plant. The resultant
RTD curve is plotted in Figure 17. The RTD calculations showed that the single column
operated with a relatively long residence time of τ=12.0 minutes. However, the
calculations showed that the column was relatively well mixed, as indicated by the low
39
13 9 5 1
14 10 6 2
15 11 7 3
16 12 8 4
Figure 16. Visualization of column mixing using a red dye and timed-sequence
images (column equipped with baffles) at one-second intervals.
Pe=1.58. In fact, the tracer distribution suggests that nearly 10% of the feed slurry exited
the cell after just one minute. Therefore, despite the long residence time, the columns
achieved a recovery of only 73.7% (two-run average) due to the large dispersion.
reconfigure the cells to operate in series as opposed to the parallel feeding arrangement
currently in use. As shown in Figure 18, the operating point defined by Pe=1.58 and
40
1.0
Single Column
0.7
Two-Column
0.6 Circuit
0.5 (Pe=4.22)
0.4
0.3
0.2
0.1
0.0
0 10 20 30 40
Time (Minutes)
Figure 17. Comparison of RTD data for single- and two-stage column circuits.
configuration. Two cells connected in series would reduce the dimensionless rate to
17. The combined recovery (R2) for the two cells in series can be estimated using
41
100
Plug-Flow
90 Pe=1.58
Recovery (%)
80
73.7
70 Perfectly
Mixed
60
54.5
50
1.97
42.7
40
0 1 2 3 4 5 6
Dimensionless Rate (k)
Figure 18. Recovery curve for projecting two- and three-stage column circuit
performance...
.
R3 = 1 − (1 − Ri ) 3 = 1 − (1 − 0.427) 3 = 81.2% [17]
The use of more than three columns may not be practical since bubbles may not be able
to rise against the increased downward flow rate passing through each column. It should
be noted that, in practice, the projected improvements are likely to slightly different from
these values since the reconfiguration may result in higher flow rates that could change
in the column plant was reconfigured so that (i) the feed slurry to columns A and B was
entirely diverted to column A and (ii) the tailings stream from column A was fed to
42
column B. The columns were sampled before and after the modification. RTD
measurements were also taken for the modified two-stage circuit. The performance
calculations and experimental results for these tests are summarized in Table 4. As
expected, the modification reduced axial dispersion by increasing Pe from 1.58 to 4.22
(see Figure 17). The modification did not substantially change the mean residence time
since twice the feed flow passed through twice the cell volume after the modification.
However, the change in feeding configuration shifted the performance for the overall
circuit from a recovery of 73.7% (two-run average) for the single column to 79.8% (two-
run average) for the two-column bank. This improvement agrees very well with the
79.3% recovery projected from the series flow calculation given in Equation [16].
The results obtained from the trial run were promising enough to move ahead
with the reconfiguration of the entire five column circuit to a cell-to-cell system. In order
Table 4. Example of RTD data for single- and two-stage column circuits.
43
to avoid excessive downward velocities in the columns, the layout was modified so that
the feed was split between the two end columns in the five cell bank and directed inward
to the center cell (Figure 19). Note that, due to piping constraints, a small amount of feed
Daily production data obtained before and after the circuit modifications are
plotted in Figure 20. Performance indicators such as average recovery and ash values are
summarized in Table 5. The daily values show substantial data scatter due to the large
variations in the feed quality from the pond reclaim operation that feeds the columns. For
this reason, weekly averages have been included on the plot so that the production trends
Feed
Clean
(a) Tails
Feed
Clean
(b) Tails
Figure 19. Column circuit layout (a) before and (b) after cell-to-cell
reconfiguration.
44
100
Circuit
90 Modified
Recovery (%)
80
70
60
0 50 100 150 200
Duration (Days)
Figure 20. Production statistics before and after the cell-to-cell column
reconfiguration.
.
can be more clearly visualized (thick solid line). Prior to the modification, the columns
produced an average recovery of 77.0% with a standard deviation of 4.3%. After the
change, the recovery improved to an average value of 81.9%, with a slightly smaller
standard deviation of 3.5%. As such, recovery improved by 4.9 percentage points. This
plant site.
45
Table 5. Production statistics before and after the full cell-to-cell column
reconfiguration.
AFTER
Average 49.72 15.26 82.37 48.82 81.85
Maximum 57.07 29.71 85.28 63.41 88.65
Minimum 42.80 9.54 78.42 34.51 71.26
Std Dev 3.13 4.46 1.23 5.56 3.48
46
4.0 SUMMARY AND CONCLUSIONS
1. Column flotation cells can produce a higher overall separation performance than
slimes into the froth product. Wash water rates in the range of 3.0-3.5 gpm/ft2 are
typically required to maintain the proper number of dilution washes needed to prevent
entrainment.
associated with froth carrying capacity. This constraint, which is very sensitive to
variations in particle size, establishes the total column cross-sectional area (column
number and diameter) required for a given application. Carrying capacities commonly
fall in the range of range of 0.06 to 0.24 tph/ft2, with an average of about 0.12 tph/ft2
for 100 mesh x 0 feed coals. The lower carrying capacities typically correspond to
4. Data obtained to date suggest that the selectivity of coal columns is largely
independent of column size and sparger design. The spargers must be capable of
dispersing air into small bubbles at gas rates typically exceeding 3.5 scfm/ft2 in order
47
to maintain acceptable coal recoveries and production rates. Sparger performance can
be measured by the parameter Sb, which represents the amount of bubble surface area
generated per unit time for a given column cross-sectional area. For optimum
5. Column cells can have both positive and negative impacts on downstream operations
in coal plants. For example, the ability of columns to removal of clay slimes and
reduce total ash content can lead to a significant increase in the filtration rate and a
lowering of the equilibrium moisture content of the clean coal filter cake. On the
other hand, difficulties associated with the increased stability of column froths may
create serious material handling problems and require the installation of deaeration
tanks or possibly the use of expensive defoaming agents. These factors must be fully
indicates that the poor recovery performance observed in some column flotation
7. The feed piping at a coal column flotation plant was modified from a parallel to a
series layout. The modification improved the recovery by nearly 5 percentage points,
48
as predicted by the mathematical projections based on residence time distributions.
This improvement increased the profitability of the column circuit at this particular
49
5.0 REFERENCES
2. Davis, V.L., Bethell, P.J., Stanley, F.L. and Luttrell, G.H., 1995. “Plant Practices
International Symposium, Chapter 21, S.K. Kawatra, Ed., SME, Inc., Littleton,
3. Dell, C.C., Bunyard, M.J., Rickelton, W.A. and Young, P.A., 1972. “Release
5. Dobby, G.S. and Finch, J.A., 1986a. “Flotation Column Scale-Up and
6. Dobby, G.S., and Finch, J.A., 1986b. "Particle Collection in Columns - Gas Rate
and Bubble Size Effects," Canadian Metallurgical Quarterly, Vol. 25 (1), pp. 9-
13.
7. Espinosa-Gomez, R., Yianatos, J.B., Finch, J.A., and Johnson, N.W., 1988.
Column Flotation '88, K.V.S. Sastry (Ed.), SME Inc., Littleton, CO, p. 143.
8. Finch, J.A., and Dobby, G.S., 1990. Column Flotation, Pergamon Press,
9. Finch, J.A., 1995. “Column Flotation: A Selected Review – Part IV: Novel
50
10. Groppo, J.G., and Parekh, B.K., 1992. "Comparison of Bubble Generation
Devices for Column Flotation of Fine Coal from Refuse," 121st Annual SME
11. Honaker, R.Q., Mohanty, M.K. and Ho, E., 1995. “Comparison of Column
Flotation Cells,” 12th Inter. Coal Preparation Conference, May 2-4, Lexington,
12. Huls, B.J., LaChance, C.D., and Dobby, G.S., 1991. "Bubble Generation
13. Manlapig, E.V., Jackson, B.R., and Harbort, G.J., 1993. “Jameson Cell Coal
Conference, Coal Prep 93, May 4-6, 1993, Lexington, Kentucky, pp. 203-219.
Blacksburg, VA.
16. Levenspiel, O., 1972. Chemical Reaction Engineering, Wiley, New York.
17. Luttrell, G.H., Stanley, F.L., and Bethell, P.J., 2004. “Operating and Design
18. Lynch, A.J., Johnson, N.W. Manlapig, E.V. and Thorne, C.G., 1981. Mineral and
20. McKay, J.D., Foot, D.G. and Shirts, M.D., 1988. “Column Flotation and Bubble
21. Nicol, S.K. and Bensley, C.N. Recent developments in fine coal preparation in
22. Rubinstein, J.B., 1995. Column Flotation: Processes, Designs and Practices,
23. Sastri, S.R.S, 1996. “Technical Note: Carrying Capacity in Flotation Columns,”
25. Wehner, J.F. and Wilhelm, R.H. Boundary conditions of flow reactors, Chemical
27. Xu, M., and Finch, J.A., 1989. "Effect of Sparger Surface Area on Bubble
pp. 1-6.
52
28. Yianatos, J.B., Finch, J.A., and LaPlante, A.R., 1988. “Selectivity in Column
29. Yoon, R.-H., Luttrell, G.H., Adel, G.T. and Mankosa, M.J., 1992. “The
53
6.0 APPENDIX
54
Appendix
55
56
57
Appendix
58
59
60